Multi-scene automatic control method of intelligent live broadcast all-in-one machine

Abstract

The application provides a multi-scene automatic control method of an intelligent live broadcast all-in-one machine, which comprises the following steps: the intelligent live broadcast all-in-one machine collects live broadcast pictures in real time through a camera, analyzes the live broadcast pictures, and determines the current live broadcast scene type; according to the live broadcast scene type, a focusing parameter library of the corresponding scene is called; when the camera is controlled to focus on the live broadcast pictures based on the focusing parameter library, the brightness distribution parameters and the color saturation parameters of the live broadcast pictures are synchronously collected; based on the brightness distribution parameters and the color saturation parameters, the adaptive light compensation of a light supplement lamp is linkage controlled; wherein the brightness adjustment range of the light supplement lamp is negatively correlated with the brightness distribution parameters in a linear relationship, and the color temperature adjustment range of the light supplement lamp is dynamically set according to the color saturation parameters. In the application, the characteristics of different scenes can be adapted, the problems of picture blurring and light imbalance can be avoided, and the live broadcast effect is improved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a multi-scene automatic control method for an intelligent live streaming all-in-one machine. Background Technology

[0002] With the rapid development of the live streaming industry, intelligent live streaming all-in-one machines, which integrate cameras, fill lights, and intelligent processing modules, are widely used in various scenarios such as e-commerce live streaming, educational live streaming, and entertainment live streaming. Different live streaming scenarios have significantly different requirements for image quality: for example, portrait live streaming needs to highlight the clarity of facial details and skin tone reproduction, product display live streaming needs to ensure that the main body of the product is accurately focused and the material texture is distinguishable, and scene-based live streaming needs to take into account both the overall environment and the highlight of the core area.

[0003] In existing technologies, the focus control and light compensation of intelligent live streaming all-in-one machines mostly use fixed parameters or rely on manual adjustment by the user. Fixed parameter mode cannot adapt to the characteristics and needs of different scenarios. For example, when using uniform focus parameters, there may be insufficient background blur and the subject not standing out enough in product display scenarios. Manual adjustment method requires a high level of user expertise, is cumbersome to operate, and is difficult to respond to scene changes in real time. Especially in the process of live streaming with rapid switching between multiple scenes, it is easy to cause problems such as blurry images and unbalanced lighting, which will affect the live streaming effect. Summary of the Invention

[0004] The main objective of this invention is to provide a multi-scenario automatic control method for an intelligent live streaming all-in-one machine, aiming to overcome the shortcomings of current methods that rely on fixed parameters or manual user adjustments, which affect the live streaming effect.

[0005] To achieve the above objectives, the present invention provides a multi-scenario automatic control method for an intelligent live streaming all-in-one machine, comprising the following steps:

[0006] The intelligent live streaming all-in-one machine captures live streaming footage in real time through a camera, analyzes the footage, and determines the type of the current live streaming scene.

[0007] According to the live streaming scenario type, the corresponding focus parameter library is called; the focus parameter library pre-stores the focus area weight matrix, focus speed coefficient and focus accuracy threshold for different live streaming scenarios.

[0008] When controlling the camera to focus on the live broadcast based on the aforementioned focus parameter library, the brightness distribution parameters and color saturation parameters of the live broadcast are collected simultaneously.

[0009] Based on the brightness distribution parameters and color saturation parameters, the supplementary light is controlled to perform adaptive light compensation; wherein, the brightness adjustment range of the supplementary light is negatively correlated with the brightness distribution parameters, and the color temperature adjustment range of the supplementary light is dynamically set according to the color saturation parameters.

[0010] Furthermore, the live stream footage is analyzed to determine the current live stream scenario type, including:

[0011] The live stream is analyzed at the frame level to extract scene feature parameters and match them with a preset scene feature library to automatically determine the current live stream scene type. The scene feature parameters include the complexity of the main body outline, the density of the background texture, and the motion trajectory of the dynamic target. The preset scene feature library includes three types of scene feature templates: portrait live stream scene, product display scene, and scenario-based live stream scene.

[0012] Furthermore, in the portrait live streaming scene focus area weight matrix, the weight value of the portrait face area is set to 0.7-0.9; in the product display scene focus area weight matrix, the weight value of the product main body area is set to 0.8-0.95; and in the scene-based live streaming scene focus area weight matrix, the weight value of the core scene area is set to 0.6-0.8. The camera is controlled to complete focusing based on the parameters in the focus parameter library.

[0013] Furthermore, the brightness distribution parameter is obtained by dividing the live broadcast screen into a 16×16 pixel grid and calculating the average brightness value of each grid, and the color saturation parameter is calculated using the average value of the S channel in the HSV color space.

[0014] When the average value of the S channel is 30%-50%, the color temperature adjustment range of the fill light is set to 3000K-5000K. When the average value of the S channel is higher than 50%, the color temperature adjustment range of the fill light is set to 4000K-6500K.

[0015] Furthermore, the focus region weight matrix in the focus parameter library employs a dynamic correction mechanism:

[0016] When the offset of the main body area exceeds the first preset ratio of the diagonal of the image in multiple consecutive frames, the local adjustment of the weight matrix is ​​automatically triggered. The weight values ​​of the adjacent areas in the offset direction are increased by a preset weight value, while the weight value of the original main body area remains unchanged.

[0017] The focusing speed coefficient changes stepwise with the average absolute value of the acceleration of the dynamic target's trajectory in the scene. When the absolute value of the acceleration is lower than the acceleration threshold, the focusing speed coefficient maintains the baseline value. When the absolute value of the acceleration is greater than the acceleration threshold, the focusing speed coefficient increases by a second preset ratio.

[0018] Furthermore, the adaptive lighting compensation also includes a scene migration compensation mechanism:

[0019] When a live scene type switch is detected, a transition compensation coefficient is calculated based on the brightness decay curve of the last first number of frames of the previous scene and the color saturation gradient of the first second number of frames of the new scene. The brightness of the fill light is then adjusted based on the transition compensation coefficient when the live scene type switch occurs.

[0020] Furthermore, during scene type switching, the fill light color temperature adjustment range adopts a gradual blending mechanism, specifically including:

[0021] Determine the reference color temperature range for the previous scene and the target color temperature range for the new scene respectively: the reference color temperature range is determined based on the color saturation parameters before the previous scene switch, and the target color temperature range is determined based on the color saturation parameters of the initial frame of the new scene.

[0022] The total transition time is set, and the real-time fusion weight is calculated using the real-time time during the transition process as a variable. The real-time fusion weight changes linearly with the transition time.

[0023] Based on real-time fusion weights, the reference color temperature range and the target color temperature range are fused and calculated to obtain the real-time color temperature adjustment range during the transition period.

[0024] Furthermore, the method also includes:

[0025] The intelligent live streaming all-in-one machine collects current network transmission parameters in real time to determine the network transmission level;

[0026] The image quality control parameters are adjusted based on the network transmission level and the type of live streaming scene; the image quality control parameters include focus accuracy and fill light parameters.

[0027] The network transmission parameters are updated every preset time, and the image quality control parameters are adjusted synchronously.

[0028] Furthermore, the method also includes:

[0029] The intelligent live streaming all-in-one machine presets the device calibration cycle. When calibration is started, it controls the camera to capture the standard color card image and the standard clarity test image, and at the same time controls the fill light to output light under different brightness and color temperature.

[0030] Calculate the sharpness deviation value after the camera is focused, and the deviation between the actual brightness and color temperature of the fill light and the nominal value at each level;

[0031] Based on the sharpness deviation value, adjust the focus accuracy threshold and focus speed coefficient; based on the deviation between the actual brightness and color temperature and the nominal value, adjust the brightness adjustment coefficient and color temperature adjustment range.

[0032] The present invention also provides a multi-scene automatic control device for an intelligent live streaming all-in-one machine, comprising:

[0033] The judgment unit is used to capture live footage in real time through the camera, analyze the live footage, and determine the current live scene type;

[0034] The calling unit is used to call the corresponding focus parameter library according to the live streaming scene type; the focus parameter library pre-stores the focus area weight matrix, focus speed coefficient and focus accuracy threshold for different live streaming scenes.

[0035] The focusing unit is used to simultaneously collect the brightness distribution parameters and color saturation parameters of the live broadcast image when controlling the camera to focus on the live broadcast image based on the focusing parameter library.

[0036] The control unit is used to control the fill light to perform adaptive light compensation based on the brightness distribution parameters and color saturation parameters; wherein the brightness adjustment range of the fill light is negatively correlated with the brightness distribution parameters, and the color temperature adjustment range of the fill light is dynamically set according to the color saturation parameters.

[0037] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0038] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods described above.

[0039] The present invention provides a multi-scene automatic control method for an intelligent live streaming all-in-one machine, comprising: the intelligent live streaming all-in-one machine acquiring live streaming images in real time through a camera, analyzing the live streaming images, and determining the current live streaming scene type; according to the live streaming scene type, calling the corresponding scene's focus parameter library; the focus parameter library pre-stores focus area weight matrices, focus speed coefficients, and focus accuracy thresholds for different live streaming scenes; when controlling the camera to focus on the live streaming image based on the focus parameter library, simultaneously acquiring the brightness distribution parameters and color saturation parameters of the live streaming image; based on the brightness distribution parameters and color saturation parameters, linking and controlling the fill light to perform adaptive light compensation; wherein, the brightness adjustment range of the fill light has a negative linear correlation with the brightness distribution parameters, and the color temperature adjustment range of the fill light is dynamically set according to the color saturation parameters. In this invention, by analyzing the live streaming image and determining the current live streaming scene type, the focus area weight matrix, focus speed coefficient, and focus accuracy threshold for different live streaming scenes are called to focus the live streaming image, adapting to the characteristic requirements of different scenes; at the same time, linking and controlling the fill light to perform adaptive light compensation avoids problems such as image blurring and light imbalance, thereby improving the live streaming effect. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the steps of a multi-scenario automatic control method for an intelligent live streaming all-in-one machine in one embodiment of the present invention;

[0041] Figure 2 This is a structural block diagram of a multi-scene automatic control device for an intelligent live streaming all-in-one machine according to an embodiment of the present invention;

[0042] Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0043] The implementation, functional features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] Reference Figure 1 One embodiment of the present invention provides a multi-scenario automatic control method for an intelligent live streaming all-in-one machine, comprising the following steps:

[0046] Step S1: The intelligent live streaming all-in-one machine collects live streaming footage in real time through the camera, analyzes the live streaming footage, and determines the current live streaming scene type.

[0047] Step S2: According to the live streaming scene type, call the focus parameter library of the corresponding scene; the focus parameter library pre-stores the focus area weight matrix, focus speed coefficient and focus accuracy threshold for different live streaming scenes.

[0048] Step S3: When controlling the camera to focus on the live broadcast image based on the focus parameter library, simultaneously collect the brightness distribution parameters and color saturation parameters of the live broadcast image.

[0049] Step S4: Based on the brightness distribution parameters and color saturation parameters, the supplementary light is controlled to perform adaptive light compensation; wherein, the brightness adjustment range of the supplementary light is negatively correlated with the brightness distribution parameters, and the color temperature adjustment range of the supplementary light is dynamically set according to the color saturation parameters.

[0050] It is particularly important to note that all technical steps, algorithm applications, and parameter settings in the technical solution of this application have clear technical objectives and application value. They do not utilize complex steps and algorithmic formulas to achieve simple functions. To provide detailed explanations of each step and avoid ambiguity, some conventional algorithms are used for illustration. However, this does not mean that the algorithms and technical features listed herein are the only way to implement the technical solution of this application, nor is it intended to limit the scope of protection of this application. This application is not a combination or stacking of the listed algorithms and technical features; its essence is to exemplify the implementation methods of this application to fully explain it. It does not pursue formal complexity by adding meaningless technical steps, nor does it involve the accumulation of technologies divorced from practical needs; it conforms to the conventional logic of technical improvement and design.

[0051] In this embodiment, as described in step S1 above, the aim is to achieve intelligent recognition of live streaming scenarios, providing a basis for subsequent control strategies. Specifically, the intelligent live streaming all-in-one machine utilizes its built-in camera component to continuously capture dynamic image data at a preset frame rate. After acquiring the real-time image, the image analysis module is activated to perform in-depth analysis of the image content. This analysis process typically includes two core steps: feature extraction and pattern matching. First, key information that can characterize scene attributes is extracted from the image, such as the morphological features of the main object, the texture distribution of the background environment, and the motion features of dynamic elements. Then, these extracted feature information are compared and matched with a pre-built scene feature database. The feature similarity is calculated through an algorithm, and finally, the type of the current live streaming scene is automatically determined, such as a portrait live streaming scene, a product display scene, or a scenario-based live streaming scene. This process requires no manual intervention, can respond to scene changes in real time, and ensures the accuracy and timeliness of scene type judgment.

[0052] As described in step S2 above, the core is to establish the correspondence between scene types and focus control parameters, thereby achieving scene-specific adaptation of the focus strategy. Once the current live streaming scene type is determined, a parameter retrieval mechanism is triggered, retrieving a complete set of focus parameters matching the scene type from a pre-stored focus parameter library. This focus parameter library is pre-built based on extensive experimental data and scene characteristic analysis, storing three sets of key parameters for different live streaming scenes: a focus area weight matrix, used to define the priority weight of different areas in the image during focus calculation; a focus speed coefficient, used to control the response speed of the camera's focus motor; and a focus accuracy threshold, used to set the criteria for judging focus clarity. By calling the specific parameters for the corresponding scene, it can be ensured that the camera's focusing behavior accurately matches the image presentation requirements of that scene, laying the foundation for obtaining a clear image of the core area.

[0053] As described in step S3 above, the synchronous coordination of focus control and image parameter acquisition is achieved, providing data support for light compensation. During the process of calling the focus parameters to control the camera to perform the focusing operation, the image acquisition and analysis module simultaneously initiates the detection mechanism of image optical parameters. Specifically, it calculates and acquires the brightness distribution parameters of the live image in real time. These parameters reflect the brightness differences in different areas of the image and the overall brightness level. Simultaneously, through color space conversion and analysis algorithms, it extracts the color saturation parameters of the image, which characterize the vividness and richness of the colors. The synchronous acquisition of these two parameters with the focusing process ensures both the timeliness of parameter acquisition and the accuracy of the acquired parameters in reflecting the optical characteristics of the image under the current focusing state, providing precise data for subsequent supplementary lighting adjustments.

[0054] As described in step S4 above, an adaptive lighting optimization for the live stream is achieved by establishing a linkage mechanism between optical parameters and supplementary lighting control. Using the collected brightness distribution parameters and color saturation parameters as input, the supplementary lighting control algorithm is activated to calculate the lighting compensation strategy. Specifically, the supplementary lighting brightness adjustment employs a dynamic control method correlated with the brightness distribution parameters: when the overall image is dark or locally insufficient, the adjustment range increases as the brightness value decreases; when the image brightness is high, the adjustment range decreases accordingly, achieving precise brightness balance through this negatively correlated linear relationship. Regarding color temperature adjustment, the color temperature adjustment range is dynamically defined based on the numerical range of the color saturation parameters: for images with low color saturation, a color temperature range that enhances color expression is selected; for images with high color saturation, a color temperature range that avoids oversaturation is selected. Through this linkage control, the supplementary lighting can adaptively compensate for the optical characteristics of the image in different scenarios, ultimately achieving an optimized effect of balanced brightness and natural, realistic colors in the live stream.

[0055] In one embodiment, the live stream is analyzed to determine the current live stream scene type, including:

[0056] The live stream is analyzed at the frame level to extract scene feature parameters and match them with a preset scene feature library to automatically determine the current live stream scene type. The scene feature parameters include the complexity of the main body outline, the density of the background texture, and the motion trajectory of the dynamic target. The preset scene feature library includes three types of scene feature templates: portrait live stream scene, product display scene, and scenario-based live stream scene.

[0057] In this embodiment, firstly, frame-level analysis is performed on the live stream captured by the camera. The continuous video stream is processed frame by frame to ensure the capture of instantaneous features in each frame, thereby accurately reflecting the dynamic changes of the scene. The core of frame-level analysis is to extract three key scene feature parameters: subject contour complexity, calculated using edge detection algorithms to determine the number of edges, degree of curvature, and proportion of closed regions of the main object's contour; for example, the complexity of a person's facial contour differs significantly from the complexity of a product's shape contour; background texture density, quantified using algorithms such as gray-level co-occurrence matrix or local binary mode (LBP) to quantify the richness of texture details in the background area; for example, the solid-color background texture density is low in product display scenes, while the natural background texture density is high in scene-based live streams; and dynamic target motion trajectory, recorded through inter-frame target tracking to track the displacement path, speed, and direction changes of the main moving objects in the image; for example, the small movements of the anchor in a portrait live stream differ from the motion trajectory of holding a product in a product display.

[0058] Secondly, the extracted scene feature parameters are matched with a preset scene feature library. This preset scene feature library is built based on a large amount of sample data and contains feature templates for three typical live streaming scenarios: portrait live streaming scene template, whose feature parameters are characterized by moderate complexity of the main body outline (face, body), low background texture density, and stable and small amplitude of dynamic target movement trajectory; product display scene template, characterized by varying complexity of the main body outline (product) depending on the category (e.g., high for jewelry, medium for clothing), extremely low background texture density (mostly solid color or simple scenery), and dynamic target (product) movement trajectory mainly consisting of local rotation or translation; and scenario-based live streaming scene template, characterized by high complexity of the main body outline (e.g., stage, landscape), high background texture density, and diverse dynamic target movement trajectories (which may include complex movements of multiple people or objects).

[0059] Finally, the matching degree between the real-time extracted feature parameters and the three types of templates is calculated by feature matching algorithms (such as cosine similarity, Euclidean distance, etc.). The scene type corresponding to the template with the highest matching degree is determined as the current live broadcast scene type, realizing automatic scene recognition without manual intervention, and providing accurate basis for subsequent targeted focus and fill light control.

[0060] In one embodiment, the weight value of the face region in the portrait live streaming scene focus area weight matrix is ​​set to 0.7-0.9, the weight value of the main product region in the product display scene focus area weight matrix is ​​set to 0.8-0.95, and the weight value of the core scene region in the scene-based live streaming scene focus area weight matrix is ​​set to 0.6-0.8. The camera is controlled to focus based on the parameters in the focus parameter library.

[0061] In this embodiment, the focus area weight matrix preset in the focus parameter library for different scene types essentially defines the priority of each area in the image during focus calculation by numerically assigning weight values. Specifically, in portrait live streaming scenarios, the weight value of the facial area (typically including key feature areas such as the eyes and facial contours) is set to 0.7-0.9. This high weight range means that during the focusing process, the camera will prioritize focusing on the facial area to ensure the clarity of the anchor's facial details (such as expressions and skin texture), while other areas in the image (such as the background and limbs) are assigned lower weights, thus creating a typical portrait live streaming effect with a clear face and a moderately blurred background.

[0062] For product display scenarios, the weight value of the main product area is set to a higher range of 0.8-0.95. This is because the core requirement of product display is to highlight the product's shape, material, and details (such as the texture of clothing, the buttons of electronic products, and the luster of jewelry). The high weight setting ensures that the camera focuses its resources on the main product. Even if there are auxiliary elements such as the host's hands or the display stand in the frame, it will not affect the focus accuracy of the product area, achieving a clear and sharp display effect while weakening the auxiliary elements.

[0063] In scenario-based live streaming, the weight of core scene areas (such as concert stages, outdoor landscape core areas, and indoor themed sets) is set to 0.6-0.8. This weight range is lower than that of core areas in portrait and product scenes because scenario-based live streaming needs to balance the harmony between the core area and its surrounding environment. It is necessary to ensure the clarity of the core scene to highlight the theme, while also appropriately preserving details of the surrounding environment to present the overall atmosphere. This avoids a fragmented image caused by over-focusing on the core area. Therefore, a medium weight setting achieves a scene presentation effect that is clear in the core and harmonious with the environment.

[0064] During the actual focusing process, the camera's focusing system calculates the sharpness data of each area in the image based on the aforementioned weight matrix. It prioritizes locking the focus on the area with the highest weight value and adjusts the sharpness priority of each area according to the weight allocation ratio. Finally, it achieves accurate focusing that matches the current scene, ensuring that the presentation effect of the core content of the image meets the scene requirements.

[0065] In one embodiment, the brightness distribution parameter is obtained by dividing the live broadcast screen into a 16×16 pixel grid and calculating the average brightness value of each grid, and the color saturation parameter is calculated using the average value of the S channel in the HSV color space.

[0066] When the average value of the S channel is 30%-50%, the color temperature adjustment range of the fill light is set to 3000K-5000K. When the average value of the S channel is higher than 50%, the color temperature adjustment range of the fill light is set to 4000K-6500K.

[0067] In this embodiment, the acquisition of brightness distribution parameters employs a grid-based method for refined analysis of the live stream: the entire image is uniformly divided into a 16×16 pixel grid (256 independent grids), with each grid serving as the basic unit for brightness detection. By calculating the average brightness of all pixels within each grid, the average brightness value of the 256 grids is obtained. These values ​​collectively constitute the brightness distribution parameters characterizing the brightness distribution features of the image. This method can accurately capture local brightness differences in the image (such as highlight areas and shadow areas), avoiding the problem of local light imbalance caused by relying solely on the global average brightness, and providing data support for subsequent regional adjustments of supplementary lighting.

[0068] Regarding the acquisition of color saturation parameters, the S-channel (saturation channel) of the HSV color space is used for quantification: First, the image is converted from the RGB color space to the HSV color space, where the S-channel value directly reflects the degree of color saturation (the value range is usually 0%-100%, with higher values ​​indicating more vivid colors). By calculating the average value of all pixels in the S-channel across the entire image, the color saturation parameter, which characterizes the overall color intensity of the image, is obtained. Compared to the RGB space, the HSV space's S-channel can more intuitively and independently reflect color saturation characteristics, facilitating the establishment of a precise correlation with subsequent color temperature adjustments.

[0069] Regarding the rules for adjusting the color temperature of the fill light, the adjustment range is dynamically defined based on the color saturation parameter (i.e., the average value of the S channel): when the average value of the S channel is between 30% and 50% (medium to low saturation), the color temperature of the fill light is limited to the range of 3000K-5000K. This is because medium to low saturation images need to enhance color performance with a warm to neutral color temperature (3000K is warm yellow, and 5000K is close to natural light) to avoid cool tones causing the image colors to appear flat; when the average value of the S channel is higher than 50% (high saturation), the color temperature adjustment range is adjusted to 4000K-6500K (neutral to cool), by moderately increasing the color temperature to suppress the glare of oversaturated colors, while maintaining the brightness of the colors, and avoiding warm tones from exacerbating the visual oppressive feeling of high saturation images.

[0070] Through the above parameter calculation methods and adjustment rules, the fill light can achieve targeted color temperature control based on the actual optical characteristics of the image, ensuring that the color presentation in different saturation scenes is both realistic and natural, and meets the needs of visual comfort.

[0071] In one embodiment, the focus region weight matrix in the focus parameter library employs a dynamic correction mechanism:

[0072] When the offset of the main body area exceeds the first preset ratio of the diagonal of the image in multiple consecutive frames, the local adjustment of the weight matrix is ​​automatically triggered. The weight values ​​of the adjacent areas in the offset direction are increased by a preset weight value, while the weight value of the original main body area remains unchanged.

[0073] The focusing speed coefficient changes stepwise with the average absolute value of the acceleration of the dynamic target's trajectory in the scene. When the absolute value of the acceleration is lower than the acceleration threshold, the focusing speed coefficient maintains the baseline value. When the absolute value of the acceleration is greater than the acceleration threshold, the focusing speed coefficient increases by a second preset ratio.

[0074] In this embodiment, the core of the dynamic correction mechanism for the focus area weight matrix is ​​to adjust the weight parameters in real time based on the positional shift of the subject area: continuously monitoring the positional changes of the subject area (such as a human face or a product subject) in multiple consecutive frames, and calculating the ratio of its offset to the length of the image diagonal. When this ratio exceeds a first preset ratio (e.g., 15%), it is determined that the subject has undergone significant displacement, and a local adjustment of the weight matrix is ​​automatically triggered. The adjustment rule is: increase the preset weight value (e.g., 0.1-0.2) in the adjacent area in the direction of the subject's offset (i.e., the predicted area where the subject may move), while keeping the weight value of the original subject area unchanged. This design ensures continuous attention to the original subject area and advances the focus priority of surrounding potential areas, avoiding brief defocusing when the subject moves rapidly, and achieving predictive focus tracking of dynamic subjects.

[0075] The core of the stepped adjustment of the focus speed coefficient is to adapt the focus response speed according to the motion acceleration of the dynamic target: by analyzing continuous data of the dynamic target's motion trajectory in the scene, the absolute value of its average acceleration is calculated to determine the intensity of the target's motion. When the absolute value of acceleration is lower than the set acceleration threshold (e.g., 0.5 m / s²), it indicates that the target motion is stable, and the focus speed coefficient maintains a baseline value to ensure focus accuracy. When the absolute value of acceleration exceeds this threshold, it indicates that the target motion intensifies (such as fast-moving goods or large movements of a live streamer). At this time, the focus speed coefficient is increased by a second preset ratio (e.g., 20%-30%) to speed up the response speed of the camera's focus motor. This stepped adjustment balances focus speed and accuracy. For low-speed motion, sharpness is prioritized; for high-speed motion, tracking timeliness is prioritized to avoid image blurring caused by excessively fast movement.

[0076] Through the aforementioned dynamic correction mechanism, the focus parameter library can overcome the limitations of fixed parameters and flexibly adjust the focus strategy according to the real-time movement of the subject, maintaining the clarity and stability of the core area of ​​the image even in complex dynamic scenes.

[0077] In one embodiment, the adaptive lighting compensation further includes a scene migration compensation mechanism:

[0078] When a live scene type switch is detected, a transition compensation coefficient is calculated based on the brightness decay curve of the last first number of frames of the previous scene and the color saturation gradient of the first second number of frames of the new scene. The brightness of the fill light is then adjusted based on the transition compensation coefficient when the live scene type switch occurs.

[0079] In this embodiment, when a change in the live streaming scene type is detected (such as switching from a portrait live streaming scene to a product display scene, or from a product display scene to a contextualized live streaming scene), a scene migration compensation mechanism is immediately activated. The core of this mechanism is to analyze the optical characteristics of the image before and after the switch and calculate the transition compensation coefficient to guide the adjustment of the fill light brightness.

[0080] Specifically, the brightness data of the last number of frames (e.g., 30 frames) at the end of the previous scene is extracted, and a brightness decay curve is generated by fitting these data. This curve can reflect the brightness change trend of the previous scene at the end (e.g., gradually darkening, remaining stable, or fluctuating slightly). At the same time, the color saturation data of the second number of frames (e.g., 5 frames) at the beginning of the new scene is extracted, and the color saturation gradient is calculated. This gradient is used to characterize the rate and direction of color saturation change at the beginning of the new scene (e.g., rapid increase, slow decrease, or basically stable).

[0081] Based on the aforementioned brightness decay curve and color saturation gradient, a transition compensation coefficient is calculated using a preset algorithm (or a transition compensation coefficient for multiple scene transitions can be pre-stored in the database). This coefficient comprehensively considers the differences and changes in brightness and color saturation between the preceding and following scenes, and can quantify the magnitude, rate, and trend of the adjustment required for the fill light brightness during the switching process.

[0082] During scene transitions, the brightness adjustment of the fill lights no longer directly applies the base parameters of the new scene. Instead, it is dynamically adjusted based on a transition compensation coefficient. For example, if the brightness of the previous scene shows a decreasing trend while the initial saturation gradient of the new scene is high, the transition compensation coefficient will guide the fill lights to reduce their brightness at a slower rate before gradually adjusting to the target brightness of the new scene. This avoids screen flickering or color banding caused by sudden brightness changes. Through this mechanism, a smooth transition of light from one scene to another is achieved, ensuring the visual continuity and stability of the live stream during scene transitions.

[0083] In one embodiment, during scene type switching, the fill light color temperature adjustment range employs a gradual blending mechanism, specifically including:

[0084] Determine the reference color temperature range for the previous scene and the target color temperature range for the new scene respectively: the reference color temperature range is determined based on the color saturation parameters before the previous scene switch, and the target color temperature range is determined based on the color saturation parameters of the initial frame of the new scene.

[0085] The total transition time is set, and the real-time fusion weight is calculated using the real-time time during the transition process as a variable. The real-time fusion weight changes linearly with the transition time.

[0086] Based on real-time fusion weights, the reference color temperature range and the target color temperature range are fused and calculated to obtain the real-time color temperature adjustment range during the transition period.

[0087] In this embodiment, the first step is to determine the color temperature range. Two key ranges need to be clearly defined: first, the baseline color temperature range corresponding to the previous scene, which is determined based on the color saturation parameters before the scene switch (such as the average S-channel value of the HSV color space) (for example, if the average S-channel value of the previous scene is 35%, the baseline color temperature range is set to 3000K-5000K); second, the target color temperature range corresponding to the new scene, which is determined based on the color saturation parameters of the initial frame of the new scene (for example, if the initial average S-channel value of the new scene is 60%, the target color temperature range is set to 4000K-6500K). These two ranges represent the color temperature ranges for adaptation between the previous and new scenes, providing the basic boundaries for subsequent fusion.

[0088] Secondly, the real-time fusion weights are calculated. A preset total transition time (e.g., 0.5-1.0 seconds) is used as the time period for the color temperature to completely transition from the reference range to the target range. During the transition, real-time fusion weights are generated using the real-time elapsed time as a variable: when the transition just begins (real-time time is 0), the weight is 0, and the parameters of the reference color temperature range are used entirely; as time progresses, the weight increases linearly with the transition time; when the total transition time is reached, the weight is 1, at which point the parameters of the target color temperature range are used entirely. This linearly changing weight ensures the uniformity and predictability of the transition process.

[0089] Finally, the real-time color temperature adjustment range is calculated. Based on the aforementioned real-time fusion weights, the baseline color temperature range and the target color temperature range are dynamically fused: the minimum color temperature value of the real-time adjustment range is obtained by adding the minimum color temperature of the baseline range according to the ratio of "(1 - real-time weight)" and the minimum color temperature of the target range according to the ratio of "real-time weight"; the maximum color temperature value of the real-time adjustment range is obtained by adding the maximum color temperature of the baseline range according to the ratio of "(1 - real-time weight)" and the maximum color temperature of the target range according to the ratio of "real-time weight". For example, when transitioning to 50% of the total duration, the real-time weight is 0.5, and the minimum color temperature of the real-time adjustment range at this time is "baseline minimum color temperature × 0.5 + target minimum color temperature × 0.5", and the maximum color temperature is calculated similarly.

[0090] Through this gradual blending mechanism, the color temperature adjustment range of the fill light achieves a smooth transition between the previous scene and the new scene, ensuring a natural connection of the screen colors during scene switching and avoiding visual discomfort caused by sudden changes in color temperature.

[0091] In one embodiment, the method further includes:

[0092] The intelligent live streaming all-in-one machine collects current network transmission parameters in real time to determine the network transmission level;

[0093] The image quality control parameters are adjusted based on the network transmission level and the type of live streaming scene; the image quality control parameters include focus accuracy and fill light parameters.

[0094] The network transmission parameters are updated every preset time, and the image quality control parameters are adjusted synchronously.

[0095] In this embodiment, the intelligent live streaming all-in-one machine first monitors and collects current network transmission parameters in real time through its built-in network monitoring module. These parameters typically include key indicators such as upload bandwidth (in Mbps), network latency (in ms), and data packet loss rate (in %). These parameters are evaluated according to preset threshold ranges, and the network transmission status is divided into different levels (e.g., excellent, good, average), providing a basis for subsequent image quality adjustments.

[0096] Secondly, based on the determined network transmission level and the current live streaming scenario, image quality control parameters are adjusted accordingly. These parameters primarily include focus accuracy and lighting parameters: When the network transmission level is high (e.g., a premium network), image quality is prioritized, with focus accuracy set to the standard threshold for the current scenario (e.g., maintaining high focus accuracy in product display scenarios), and lighting parameters set to the optimal standard for color reproduction and brightness balance. When the network transmission level is low (e.g., a general network), to avoid live streaming stuttering, non-core image quality parameters are appropriately reduced. For example, the frequency of focus calculations is reduced to decrease data processing volume, or the adjustment dimensions of lighting parameters are simplified (e.g., narrowing the color temperature adjustment range), reducing data transmission pressure while maintaining basic image clarity. Simultaneously, the adjustment strategy is implemented in conjunction with scenario type differences. For instance, scenario-based live streaming has higher network bandwidth requirements, and its parameter adjustment range may be greater than that for portrait live streaming scenarios at the same network level.

[0097] Finally, network transmission parameters are continuously updated at preset time intervals (e.g., 1-2 seconds) to simultaneously assess changes in network quality and dynamically adjust image quality control parameters accordingly. If the network transmission level improves (e.g., from average to good), image quality parameters are gradually restored to enhance the picture quality; if the network transmission level decreases (e.g., from good to average), non-core parameters are quickly lowered to prioritize ensuring smooth live streaming.

[0098] Through the aforementioned dynamic adjustment mechanism, the intelligent live streaming all-in-one machine can achieve a balance between live streaming quality and transmission stability in complex and ever-changing network environments, avoiding problems such as sudden drops in image quality or live streaming interruptions caused by network fluctuations.

[0099] In one embodiment, the method further includes:

[0100] The intelligent live streaming all-in-one machine presets the device calibration cycle. When calibration is started, it controls the camera to capture the standard color card image and the standard clarity test image, and at the same time controls the fill light to output light under different brightness and color temperature.

[0101] Calculate the sharpness deviation value after the camera is focused, and the deviation between the actual brightness and color temperature of the fill light and the nominal value at each level;

[0102] Based on the sharpness deviation value, adjust the focus accuracy threshold and focus speed coefficient; based on the deviation between the actual brightness and color temperature and the nominal value, adjust the brightness adjustment coefficient and color temperature adjustment range.

[0103] In this embodiment, the first step is the calibration cycle and basic data acquisition. A fixed device calibration cycle is preset in the intelligent live streaming all-in-one machine (e.g., 30 days / cycle by default), and the calibration process is automatically started upon reaching the cycle. During the calibration phase, the camera is controlled to sequentially acquire standard color chart images (for color benchmark comparison) and standard sharpness test images (including line patterns with specific resolution); simultaneously, the supplementary light is controlled to output light at preset different brightness levels (e.g., 20% / 50% / 80% of rated power) and different color temperature levels (e.g., 3000K / 4500K / 6000K), and the actual light data corresponding to each level is acquired.

[0104] Secondly, the deviation value is calculated. For the camera, by comparing the number of identifiable lines and edge sharpness in the standard resolution test image and the preset standard image, the sharpness deviation value after focusing is calculated. This value reflects the difference between the camera's current resolution and its nominal performance. For the fill light, the difference between the actual measured brightness value and the nominal brightness value at each brightness level, and the difference between the actual measured color temperature value and the nominal color temperature value at each color temperature level (e.g., nominal 4500K, actual output 4800K, deviation 300K) are calculated.

[0105] Finally, parameter correction is performed. Control parameters are adjusted based on calculated deviation values: if the camera sharpness deviation exceeds a preset threshold (e.g., 10%), the focus accuracy threshold is lowered (to improve the focus judgment standard) and the focus speed coefficient is increased (to compensate for lens response delay) to offset the impact of decreased resolution; if the supplementary light brightness deviation exceeds a threshold (e.g., 20%), the brightness adjustment coefficient is increased proportionally to the deviation (e.g., if the deviation is 15%, the coefficient increases by 15%) to ensure that the actual output brightness meets the control command; if the color temperature deviation exceeds a threshold (e.g., 500K), a deviation compensation value is added to the original color temperature adjustment range (e.g., if the actual color temperature is 300K higher, the target color temperature range is lowered by 300K overall) to ensure the accuracy of color control.

[0106] Through the aforementioned periodic calibration and parameter correction, the equipment can continuously compensate for performance deviations caused by lens aging, light source attenuation, etc., maintain stable focusing accuracy and lighting effect over a long period of time, and extend the effective service life of the equipment.

[0107] In one embodiment, the method further includes:

[0108] Obtain relevant information during the live stream, including the sequence of changes in the number of viewers, the duration of the live stream, and the live stream account identifier;

[0109] Generate a viewer change curve based on the viewer count change sequence;

[0110] Based on the live broadcast duration, a corresponding graphic is generated; a preset character array is obtained, and the graphic is superimposed on the preset character array in a preset manner; wherein, the center of the graphic overlaps with the center of the character array, and the size ratio of the graphic to the preset character array conforms to a preset ratio;

[0111] Based on the positional relationship between the graphic and each array element in the preset character array, each character in the live account identifier is distributed and added to multiple array elements to obtain an optimized character array;

[0112] The number of people change curve is superimposed onto the optimized character array in a preset manner. In each column of the optimized character array, the array element that is farthest from the center of the curve segment in the current column is obtained as the target element. The target elements are combined to obtain the encryption key.

[0113] Extract multiple key frames from the live stream and store them in a preset folder. Then, encrypt the preset folder using the encryption key.

[0114] In this embodiment, the first step is the collection of live-streaming related information. During the live stream, the intelligent live-streaming all-in-one machine acquires three types of core related information in real time: first, the sequence of viewership changes, which is the viewership data recorded at fixed time intervals (e.g., 1 minute / time), forming a sequence of viewership that fluctuates over time; second, the live stream duration, which is the cumulative duration from the start of the live stream to the current moment (accurate to the second); and third, the live stream account identifier, which is the information used to uniquely identify the live stream entity (e.g., the streamer's account ID, the live stream room name, etc.). This information together constitutes the basic data source for key generation.

[0115] Secondly, the feature transformation of related information and the overlay of character arrays are performed. The first step is to generate a viewer count change curve based on the viewer count change sequence. Using the live stream duration as the horizontal axis and the viewer count as the vertical axis, the viewer count data at each time point is connected to form a continuous curve, intuitively reflecting the fluctuation trend of the number of viewers. The second step is to generate corresponding graphics based on the live stream duration. For example, the preset rule is to generate a circle for a live stream duration ≤ 1 hour, a rectangle for 1-3 hours, and a pentagon for more than 3 hours, with the graphic size positively correlated with the live stream duration (e.g., the side length / diameter of the graphic increases by 5mm for every additional 30 minutes). The third step is to obtain a preset fixed-specification character array (e.g., 4×4 or 5×5, with array elements being a random combination of letters and numbers). The generated graphic is then overlaid onto the character array at a preset ratio where the center of the graphic completely coincides with the center of the array, and the diameter of the graphic's outer circle is 70%-80% of the array's diagonal length, thus clarifying the spatial relationship between the graphic and the array.

[0116] Next comes the construction of the optimized character array and the generation of the encryption key. On one hand, based on the positional relationship between the graphic and the character array (such as the array elements covered by the graphic outline, the array elements corresponding to the graphic vertices, etc.), the characters of the live broadcast account identifier are scattered into the array: for example, the characters of the account ID are filled into the array elements covered by the graphic outline, and the characters of the live broadcast room name are filled into the elements corresponding to the graphic vertices, ensuring that the account identifier is not stored centrally, but is scattered in multiple array elements, forming an optimized character array containing account information; on the other hand, the curve of the number of viewers is superimposed on the optimized character array, with the center of the curve overlapping the center of the array. Then, in each column of the array, the array element farthest from the center of the curve segment in that column (i.e., the element with the largest spatial distance from the center of the curve segment in the column) is found, and these elements are combined sequentially according to the order of the columns to form the key used for encryption.

[0117] Finally, there is the encrypted storage of key footage. Multiple key frames are extracted from the live stream (such as frames with the highest image clarity, most typical scene features, and best lighting effects), and these key frames are stored in a pre-defined local folder. Then, a pre-defined encryption algorithm (such as a symmetric encryption algorithm), using the generated encryption key as the core key, is employed to encrypt the folder containing the key footage. This prevents unauthorized access and leakage of key footage, ensuring the security of the core live stream content.

[0118] In this embodiment, by combining dynamic data (number of people, duration) in the live streaming scenario with a static character array, the encryption key is strongly associated with the live streaming process. This ensures both the uniqueness and randomness of the key, and also enables the secure storage of key live streaming footage, thereby improving the security of the live streaming data.

[0119] In one embodiment, the method further includes:

[0120] Collect scene-related data during the live broadcast, including the frequency of audience interaction, the unique identification code of the live broadcast account, and the change value of the fill light color temperature when switching scenes;

[0121] Generate scene feature codes for live streaming scene types;

[0122] An interaction fluctuation graph is generated based on the frequency of audience interaction. The graph is drawn with the live broadcast duration as the horizontal axis and the interaction frequency as the vertical axis. Closed graphs are drawn for the periods when the interaction frequency exceeds the average value. The number of sides of the graph is consistent with the number of periods exceeding the average value, and the area of ​​the graph is positively correlated with the total interaction frequency.

[0123] Obtain a preset scene-character mapping table, where different scene feature codes correspond to different character sets; based on the scene types involved in the current live broadcast, extract the corresponding character sets from the mapping table to construct a dynamic character array;

[0124] The interactive undulating graphics are superimposed onto the dynamic character array according to the rules. The array elements that the graphic edges pass through are marked as feature elements. The exclusive identification code of the live broadcast account is filled into the feature elements in the odd row according to the rule of filling the ID characters into the feature elements in the even row and filling the authentication number into the feature elements in the even row, thus forming a scene-account associated array.

[0125] Calculate the mapping relationship between the color temperature change value of the fill light in each scene and the number of array columns. In each column, select the unfilled feature element corresponding to the largest color temperature change value as the key unit. Combine the key units and perform an XOR operation with the scene feature code to generate the final encryption key.

[0126] Extract multiple key frames from the live stream and store them in a preset folder. Then, encrypt the preset folder using the encryption key.

[0127] In this embodiment, the intelligent live streaming all-in-one machine captures three types of key scene-related data in real time during the live stream: audience interaction frequency (total number of comments, likes, and shares per minute, reflecting fluctuations in live stream popularity), live stream account-specific identification code (including a unique account ID and streamer authentication number, used to bind the encrypted entity's identity), and fill light color temperature change value during scene switching (recording the specific difference in fill light color temperature when switching from scene A to scene B, reflecting the optical characteristics of scene switching). This data includes both the live stream's identity information and the dynamic changes in the scene, providing foundational data support for subsequent encryption from both identity and scene dimensions.

[0128] By transforming abstract data into associative feature codes and graphics, a bridge is built between data and the encrypted carrier: Scene Feature Code Generation: Based on the live streaming scene type, the most representative optical or content features of that scene are extracted (e.g., facial parameters for portrait scenes, product area parameters for product scenes), and converted into fixed-length scene feature codes (e.g., 8-bit binary codes). The core function of the feature codes is to establish the correspondence between scene type and encrypted character set, ensuring that the encryption logic of different scenes is adapted to the characteristics of the scene.

[0129] Interactive Fluctuation Graph Generation: A coordinate system is constructed with live stream duration on the horizontal axis and audience interaction frequency on the vertical axis. First, the average interaction frequency throughout the live stream is calculated. Then, all time periods with interaction frequencies exceeding the average are selected. Based on the duration of these time periods and their corresponding interaction frequencies, closed graphs are drawn. The number of sides of the graph strictly matches the number of time periods exceeding the average (e.g., a quadrilateral is drawn for four times the average). The area of ​​the graph is positively correlated with the total interaction frequency throughout the live stream (the higher the total interaction frequency, the larger the graph area). This graph transforms abstract interactive data into a concrete spatial graphic, providing a spatial basis for the subsequent feature labeling of the character array.

[0130] Dynamic character array construction: The preset scene-character mapping table is called, which predefines the exclusive character set corresponding to different scene feature codes (such as the 26 English letters for portrait scene feature codes and the 0-9 digits for product scene feature codes). Based on the current live broadcast scene type and the corresponding scene feature code, the matching character set is extracted from the mapping table, and the number of array rows is determined based on the number of scene switching during the live broadcast (such as building a 3-row array if the scene is switched 3 times). Finally, a dynamic character array strongly associated with the scene (rather than a fixed-size array) is formed to ensure that the array character set is adapted to the scene characteristics.

[0131] Feature element marking and identity filling: The generated interactive undulating graphics are superimposed onto the dynamic character array according to the rule that the center of gravity of the graphics completely coincides with the center of gravity of the array, and the ratio of the perimeter of the graphics to the perimeter of the array is 1:2. The array elements that the edges of the graphics directly pass through are marked and defined as feature elements (i.e., core elements strongly associated with the interactive data). Then, according to the rule of filling the odd-numbered rows of feature elements with account ID characters and the even-numbered rows with anchor authentication number characters, the live broadcast account's unique identifier is distributed and filled into the feature elements to form a scene-account associated array. The above filling method avoids centralized storage of identity information and improves the uniqueness of the array and the security of the encryption foundation through multiple associations.

[0132] Key unit filtering: First, calculate the color temperature change value of the fill light when switching between scenes, and then establish a mapping relationship between the color temperature change value and the number of columns in the dynamic character array (e.g., the first column corresponds to the color temperature change value of the first scene switch, and the second column corresponds to the color temperature change value of the second scene switch). In each column, filter out the unfilled feature elements when the color temperature change value of the corresponding scene is the largest (i.e., the elements that have not been filled with account information but are associated with the feature of the largest color temperature change), and define these elements as key units.

[0133] Final key generation: Combine all key units into an initial key; then perform an XOR operation between the initial key and the previously generated scene feature code to obtain the final encryption key with three attributes: scene feature, interaction feature, and identity feature.

[0134] Finally, multiple key frames are extracted from the live stream (usually the frames with the highest image clarity, most typical scene features, and best lighting effects, such as product detail frames in a product display scene or clear facial frames in a portrait scene). These key frames are then stored in a pre-defined local folder (the folder name can combine the live stream account ID and scene feature code for easy identification later). Finally, a pre-defined encryption algorithm is used, with the generated final encryption key as the core key, to encrypt the folder storing the key frames. Only authorized users with the correct key can decrypt and access the files, ensuring that the key frames of the live stream are not obtained or tampered with without authorization.

[0135] In the above embodiments, this application incorporates some existing algorithms and technical features for explanation and description to make the specification more detailed, clear, and complete, thus complying with the provisions of the Patent Law. However, this is not achieved by using a series of complex steps and algorithmic formulas, nor by complicating the technical solution, nor by combining or stacking conventional or simple features. The existing algorithms and technical features listed are for the purpose of disclosing the specific implementation methods of each step of this application (not to limit this application) and to avoid situations where this application cannot be implemented.

[0136] Reference Figure 2 In another embodiment of the present invention, a multi-scene automatic control device for an intelligent live streaming all-in-one machine is also provided, comprising:

[0137] The judgment unit is used to capture live footage in real time through the camera, analyze the live footage, and determine the current live scene type;

[0138] The calling unit is used to call the corresponding focus parameter library according to the live streaming scene type; the focus parameter library pre-stores the focus area weight matrix, focus speed coefficient and focus accuracy threshold for different live streaming scenes.

[0139] The focusing unit is used to simultaneously collect the brightness distribution parameters and color saturation parameters of the live broadcast image when controlling the camera to focus on the live broadcast image based on the focusing parameter library.

[0140] The control unit is used to control the fill light to perform adaptive light compensation based on the brightness distribution parameters and color saturation parameters; wherein the brightness adjustment range of the fill light is negatively correlated with the brightness distribution parameters, and the color temperature adjustment range of the fill light is dynamically set according to the color saturation parameters.

[0141] In this embodiment, the specific implementation of each unit in the above device embodiment is described in the above method embodiment, and will not be repeated here.

[0142] Reference Figure 3 This invention also provides a computer device, which can be a server, and its internal structure can be as follows: Figure 3 As shown, the computer device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the data corresponding to this embodiment. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0143] Those skilled in the art will understand that Figure 3 The structures shown are merely block diagrams of some structures related to the present invention and do not constitute a limitation on the computer devices on which the present invention is applied.

[0144] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0145] In summary, the multi-scene automatic control method for the intelligent live streaming all-in-one machine provided in this embodiment of the invention includes: the intelligent live streaming all-in-one machine acquiring live streaming images in real time through a camera, analyzing the live streaming images, and determining the current live streaming scene type; calling the corresponding scene's focus parameter library according to the live streaming scene type; the focus parameter library pre-stores focus area weight matrices, focus speed coefficients, and focus accuracy thresholds for different live streaming scenes; when controlling the camera to focus on the live streaming image based on the focus parameter library, simultaneously acquiring the brightness distribution parameters and color saturation parameters of the live streaming image; and based on the brightness distribution parameters and color saturation parameters, linking and controlling the fill light to perform adaptive light compensation; wherein, the brightness adjustment range of the fill light is negatively correlated with the brightness distribution parameters, and the color temperature adjustment range of the fill light is dynamically set according to the color saturation parameters. In this invention, by analyzing the live broadcast image, the current live broadcast scene type is determined, and then the focus area weight matrix, focus speed coefficient and focus accuracy threshold under different live broadcast scenes are called to focus the live broadcast image to adapt to the characteristic requirements of different scenes; at the same time, the fill light is linked to perform adaptive light compensation to avoid problems such as image blurring and light imbalance, thereby improving the live broadcast effect.

[0146] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0147] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0148] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A multi-scene automatic control method of an intelligent live broadcast all-in-one machine, characterized in that, The method comprises the following steps: The intelligent live broadcast all-in-one machine collects live broadcast pictures in real time through a camera, analyzes the live broadcast pictures, and determines the current live broadcast scene type; According to the live broadcast scene type, a focus parameter library corresponding to the scene is called; The focus parameter library pre-stores a focus area weight matrix, a focus speed coefficient, and a focus accuracy threshold value under different live broadcast scenes; The focus area weight matrix in the focus parameter library adopts a dynamic correction mechanism: when the subject area offset in continuous multiple frames is detected to exceed a first preset proportion of the picture diagonal, the weight matrix is automatically triggered for local adjustment, wherein the weight value of the adjacent area in the offset direction is increased by a preset weight value, and the original subject area weight value remains unchanged; The focus speed coefficient changes in a stepwise manner with the average acceleration absolute value of the scene dynamic target motion trajectory, and when the acceleration absolute value is lower than an acceleration threshold value, the focus speed coefficient maintains a reference value, and when the acceleration absolute value is greater than the acceleration threshold value, the focus speed coefficient is increased by a second preset proportion; When the camera is controlled based on the focus parameter library to focus on the live broadcast picture, the brightness distribution parameter and the color saturation parameter of the live broadcast picture are synchronously collected; Based on the brightness distribution parameter and the color saturation parameter, the light compensation lamp is controlled to perform adaptive light compensation; wherein the light compensation lamp brightness adjustment amplitude and the brightness distribution parameter are negatively correlated in a linear relationship, and the light compensation lamp color temperature adjustment range is dynamically set according to the color saturation parameter.

2. The multi-scene automatic control method of the intelligent live broadcast all-in-one machine according to claim 1, characterized in that, The live broadcast picture is analyzed, and the current live broadcast scene type is determined, comprising: Frame-level analysis is performed on the live broadcast picture, scene feature parameters in the picture are extracted, and the scene feature parameters are matched with a preset scene feature library to automatically determine the current live broadcast scene type; the scene feature parameters include subject contour complexity, background texture density, and dynamic target motion trajectory; the preset scene feature library includes three types of scene feature templates, namely, portrait live broadcast scene, commodity display scene, and scene live broadcast scene.

3. The multi-scene automatic control method of the intelligent live broadcast all-in-one machine according to claim 2, characterized in that, Among them, The portrait live broadcast scene sets the weight value of the portrait face area in the focus area weight matrix to 0.7-0.9, the commodity display scene sets the weight value of the commodity main area in the focus area weight matrix to 0.8-0.95, and the scene live broadcast scene sets the weight value of the core scene area in the focus area weight matrix to 0.6-0.

8. The camera is controlled based on the parameters in the focus parameter library to complete focusing.

4. The multi-scene automatic control method of the intelligent live broadcast all-in-one machine according to claim 1, characterized in that, Among them, The brightness distribution parameter is obtained by dividing the live broadcast picture into 16x16 pixel grids and calculating the average brightness value of each grid, and the color saturation parameter is calculated using the S channel mean value in the HSV color space; When the S channel mean value is 30%-50%, the light compensation lamp color temperature adjustment range is set to 3000K-5000K, and when the S channel mean value is higher than 50%, the light compensation lamp color temperature adjustment range is set to 4000K-6500K.

5. The multi-scene automatic control method of the intelligent live broadcast all-in-one machine according to claim 1, characterized in that, The adaptive light compensation also includes a scene transition compensation mechanism: When the live broadcast scene type is detected to switch, a transition compensation coefficient is calculated based on the brightness decay curve of the last first number of frames of the previous scene and the color saturation gradient of the initial second number of frames of the new scene, and the light compensation lamp brightness during the live broadcast scene type switching is adjusted based on the transition compensation coefficient.

6. The multi-scene automatic control method of the intelligent live broadcast all-in-one machine according to claim 5, characterized in that, During the scene type switching, the color temperature adjustment range of the fill light adopts a gradual fusion mechanism, specifically including: respectively determining a reference color temperature interval corresponding to the previous scene and a target color temperature interval corresponding to the new scene: wherein the reference color temperature interval is determined according to the color saturation parameter before the previous scene switching, and the target color temperature interval is determined according to the color saturation parameter of the initial frame of the new scene; setting a total transition time length, taking the real-time time in the transition process as a variable, calculating the real-time fusion weight, and the real-time fusion weight changes linearly with the lapse of the transition time; based on the real-time fusion weight, the reference color temperature interval and the target color temperature interval are fused and calculated to obtain the real-time color temperature adjustment interval during the transition.

7. The multi-scene automatic control method of the intelligent live broadcast all-in-one machine according to claim 1, characterized in that, The method further includes: The intelligent live broadcast all-in-one machine collects the current network transmission parameter in real time to determine the network transmission level; adjusting the picture quality control parameter based on the network transmission level and the live broadcast scene type; the picture quality control parameter includes focusing accuracy, fill light parameter; the network transmission parameter is updated once every preset time, and the picture quality control parameter is adjusted synchronously.

8. The multi-scene automatic control method of the intelligent live broadcast all-in-one machine according to claim 5, characterized in that, The method further includes: presetting a device calibration period in the intelligent live broadcast all-in-one machine, when the calibration is started, controlling the camera to collect the standard color card picture and the standard definition test chart, and simultaneously controlling the fill light to output light at different brightness and different color temperature; calculating the definition deviation value after the camera focusing, and the actual brightness, color temperature and nominal value deviation of the fill light at each gear; based on the definition deviation value, adjusting the focusing accuracy threshold and the focusing speed coefficient; according to the actual brightness, color temperature and nominal value deviation, adjusting the brightness adjustment coefficient and the color temperature adjustment range.

Images